Welcome to MCP Nest

MCP Nest is your gateway to running Model Context Protocol (MCP) servers in the cloud. No more local setup hassles, dependency issues, or keeping your computer running 24/7. This guide walks you through what MCP Nest does and how it works in this app.

What is MCP Nest?

MCP Nest takes your mcpServers configuration and runs those servers in the cloud, then exposes a single authenticated MCP endpoint that your AI assistant can connect to. Whether you’re using Claude’s Remote Connectors, ChatGPT, or any MCP‑compatible client, your tools are available from anywhere.

Behind the scenes, MCP Nest:

• Stores your MCP configuration securely
• Deploys servers to our infrastructure
• Aggregates all tools behind one endpoint (/mcp)
• Authenticates requests with a per‑user token
• Optionally filters the tool list to hide disabled tools you’ve toggled in the dashboard

Prerequisites

Before you begin, you’ll need:

• An MCP Nest account (click “Log in” from the homepage)
• Your MCP server configuration (the mcpServers JSON)
• API keys for any services your MCP servers require

Step 1: Understanding Your MCP Configuration

Your MCP configuration file typically looks like this:

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-github"
      ],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "gho_...",
        "GITHUB_TOOLSETS": "repos,issues,pull_requests"
      }
    }
  }
}

This configuration defines:

• Server name: github - A unique identifier for this server
• Command: The executable to run (npx, uvx, node, etc.)
• Args: Command-line arguments
• Environment variables: API keys and configuration options
• Transport: Use {"type": "stdio"} for most servers

Step 2: Preparing Your Configuration

1. Gather your API keys: Make sure you have all necessary API keys for your MCP servers
2. Test locally (optional): If you have the tools installed, test your configuration locally first
3. Remove sensitive defaults: Don’t include actual API keys in the configuration file you’ll upload

Step 3: Save and Deploy

1. Log into your MCP Nest dashboard
2. Open “Config”
3. Paste your mcpServers JSON and click Save
4. Click “Deploy” — this will build and roll out your servers in the cloud
5. Watch the status; when it reaches “deployed”, you’re ready to connect

You can view live logs at any time from the “Logs” page.

Note: Deployments are per‑user; only your authenticated requests will be routed to your latest active deployment.

Step 4: Connect your AI assistant

Open the Dashboard. You’ll see two read‑only fields:

• MCP Endpoint: the base /mcp path on this app
• Authenticated URL: the same endpoint with your personal token appended (for convenience)

Copy the “Authenticated URL” and use it in your client.

Claude Desktop (Remote Connectors)

Add this to your Claude Desktop configuration:

{
  "mcpServers": {
    "mcp-nest": {
      "transport": "http",
      "url": "YOUR-AUTHENTICATED-URL-HERE"
    }
  }
}

ChatGPT and other clients

Point your MCP client to the same authenticated URL using the HTTP transport.

Step 5: Managing Your Deployments

Viewing Logs

Check server logs from the “Logs” page to diagnose startup and runtime issues.

Updating Configurations

To update a deployed configuration:

1. Open “Config”
2. Edit your JSON
3. Save and click “Deploy” again — we’ll roll out the change to your active deployment

Monitoring Usage

You can see aggregated tools and counts on the Dashboard once servers are refreshed.

Common Use Cases

Running Multiple MCP Servers

MCP Nest aggregates multiple servers into a single endpoint:

{
  "mcpServers": {
    "github": { ... },
    "perplexity": { ... },
    "weather": { ... }
  }
}

All these servers become available through one connection to your AI assistant.

Fine‑grained control over tools

From the Dashboard you can disable individual tools (per user). Disabled tools are removed from the tools/list response so your client only sees what you’ve allowed. See the “How to disable individual MCP tools” guide for details.

Troubleshooting

Server Not Responding

• Check the logs for error messages
• Verify your API keys are correct
• Ensure the MCP server package exists and is correctly specified

Connection Issues

• Verify your endpoint URL is correct
• Check your authentication token
• Ensure your AI assistant supports HTTP transport

Performance Issues

• Review server logs for bottlenecks
• If a particular server is slow to respond, try disabling heavy tools to reduce list size

Best Practices

1. Security: Never commit API keys to version control
2. Testing: Test configurations locally when possible
3. Monitoring: Regularly check logs for errors
4. Updates: Keep your MCP server packages up to date
5. Documentation: Document your server configurations for team members

Next Steps

Now that you have MCP Nest set up:

• Explore your aggregated tools on the Dashboard
• Learn how to disable individual MCP tools

Need Help?

• Check logs first — they usually pinpoint issues
• If something looks off in the UI, hit “Refresh” on the Dashboard to reload tool lists

Welcome to the nest! 🪹